@Override
protected void init(AbstractStochasticCachingDiffFunction func) {
sList = new ArrayList<double[]>();
yList = new ArrayList<double[]>();
dir = new double[func.domainDimension()];
}
@Override
protected void takeStep(AbstractStochasticCachingDiffFunction dfunction) {
try {
computeDir(dir, newGrad);
} catch (SQNMinimizer.SurpriseConvergence s) {
clearStuff();
}
double thisGain = gain * gainSchedule(k, 5 * numBatches);
for (int i = 0; i < x.length; i++) {
newX[i] = x[i] + thisGain * dir[i];
}
// Get a new pair...
say(" A ");
if (M > 0 && sList.size() == M || sList.size() == M) {
s = sList.remove(0);
y = yList.remove(0);
} else {
s = new double[x.length];
y = new double[x.length];
}
dfunction.recalculatePrevBatch = true;
System.arraycopy(dfunction.derivativeAt(newX, bSize), 0, y, 0, grad.length);
// compute s_k, y_k
ro = 0;
for (int i = 0; i < x.length; i++) {
s[i] = newX[i] - x[i];
y[i] = y[i] - newGrad[i] + lambda * s[i];
ro += s[i] * y[i];
}
ro = 1.0 / ro;
sList.add(s);
yList.add(y);
roList.add(ro);
}
